Updating Pediatric Reference Ranges at Seattle Children’s Hospital Using an Indirect Approach

Abstract Background The aim of this project was to evaluate the accuracy of historical reference ranges at our institution. Our population of interest is children (<18 years old) within a large city. Ideally, reference intervals are determined using the direct method by performing studies on healthy volunteers. However, this approach has many challenges including being resource intensive, and, particularly in the pediatric population, difficulty in obtaining the required number of healthy volunteers for every age range. Given these challenges, a second approach, the indirect approach, has also been utilized to update reference ranges. This approach uses retrospective data analysis on data that was not specifically collected for evaluation of a reference range. Abnormal values are filtered out using a variety of methods. Methods For our project we pulled all outpatient data between 10/3/2020 and 06/30/2023 for the following analytes: albumin, ALP, AST, ALT, BUN, calcium, chloride, CO2, conjugated bilirubin, delta bilirubin, potassium, sodium, total bilirubin, total protein, and unconjugated bilirubin. We limited each patient to their first result for each analyte on the basis that analytes being measured more than once are more likely to be abnormal, and to limit bias. We also pulled the age (in days) and legal sex for each patient. The Tukey method was used to exclude outliers on the assumption that most data points would be normal, so any outliers would represent abnormal values from potentially not healthy patients. For each analyte and every age range the frequency of each value was plotted and the percentage of values within the current reference range, 2.5th percentile, and 97.5th percentile were calculated. A reference range was considered accurately representative of the normal population at our institution if the percentage of values within the reference range was greater than or equal to 80%. Results The “in range” percentage for albumin, ALT, BUN, calcium, chloride, conjugated bilirubin, delta bilirubin, creatinine, and total bilirubin was greater than 80% for all age ranges tested. Potassium, sodium, and unconjugated bilirubin each had one age range where the percent “in range” was less than 80% (2 days – 3 months, < 1 week, and < 1 month respectively), however the sample number for these age ranges was considered too low to make any significant conclusions (n=273, n=11, and n=21 respectively). ALP, AST, CO2, and total protein each had at least one age range where the percent “in range” was less than 80% and a substantial number of data points was available. In these cases, we proposed changing the reference interval to the calculated 2.5th and 97.5th percentiles based on our retrospective data analysis or verifying an externally validated pediatric reference interval from the CALIPER study.